WO2019076110A1 - 图像信息处理方法及装置 - Google Patents

图像信息处理方法及装置 Download PDF

Info

Publication number
WO2019076110A1
WO2019076110A1 PCT/CN2018/099598 CN2018099598W WO2019076110A1 WO 2019076110 A1 WO2019076110 A1 WO 2019076110A1 CN 2018099598 W CN2018099598 W CN 2018099598W WO 2019076110 A1 WO2019076110 A1 WO 2019076110A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
sub
information
extracting
size
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2018/099598
Other languages
English (en)
French (fr)
Inventor
高灵波
刘文峰
黄钦
陈恒
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Gree Electric Appliances Inc of Zhuhai
Gree Wuhan Electric Appliances Co Ltd
Original Assignee
Gree Electric Appliances Inc of Zhuhai
Gree Wuhan Electric Appliances Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Gree Electric Appliances Inc of Zhuhai, Gree Wuhan Electric Appliances Co Ltd filed Critical Gree Electric Appliances Inc of Zhuhai
Publication of WO2019076110A1 publication Critical patent/WO2019076110A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
    • G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
    • G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
    • G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]

Definitions

  • the present application relates to the field of image processing, and in particular to an image information processing method and apparatus.
  • the embodiment of the present application provides an image information processing method and apparatus, so as to at least solve the technical problem of high hardware cost consumption caused by increasing the convolution depth.
  • an image information processing method including: acquiring an image to be processed; converting the image to be processed into at least one sub image; and extracting at least one of the at least one size according to the size of the at least one sub image. Image information of the sub image.
  • the extracting the image information of the at least one sub-image according to the size of the at least one sub-image includes: determining whether the size of the at least one sub-image is smaller than a preset threshold; if the size of the at least one sub-image is smaller than the foregoing Presetting a threshold value, extracting contour information of the at least one sub-image; extracting, when the size of the at least one sub-image is greater than or equal to the preset threshold, detail information of the at least one sub-image; wherein the image information includes the contour information and the foregoing Details.
  • the extracting the detailed information of the at least one sub-image includes: extracting image information of the at least one sub-image by using the same type of two-dimensional sampling; and extracting contour information of the at least one sub-image from the image information to obtain detailed information. .
  • the method further includes: performing convolution operation on the image information of the at least one sub-image to output at least one convolution result ;
  • the respective convolution results are summed to output the aggregated result.
  • performing convolution operation on the image information of the at least one sub-image to output the at least one convolution result comprises: performing convolution operations on the image information of the at least one sub-image by the plurality of convolution operation modules, respectively The at least one convolution result described above is output.
  • an image information processing apparatus including: an acquiring unit, configured to acquire an image to be processed; and a converting unit, configured to convert the image to be processed into at least one sub image; and extract And a unit, configured to extract image information of the at least one sub-image according to the size of the at least one sub-image.
  • the extracting unit includes: a determining module, configured to determine whether the size of the at least one sub-image is smaller than a preset threshold; and an extracting module, configured to extract the at least if the size of the at least one sub-image is smaller than the preset threshold The contour information of the sub-image; if the size of the at least one sub-image is greater than or equal to the preset threshold, extracting the detail information of the at least one sub-image; wherein the image information includes the contour information and the detailed information.
  • an image information processing apparatus including: an image generator for acquiring an image to be processed; converting the image to be processed into at least one sub image; a processor, the above processing
  • the program runs the program, wherein the program output performs the following processing steps on the data output from the image generator: extracting image information of the at least one sub-image according to the size of the at least one sub-image.
  • the processor is configured to: according to the size of the at least one sub-image, extract image information of the at least one sub-image: determining whether a size of the at least one sub-image is smaller than a preset threshold; Extracting, according to the preset threshold, a size of the at least one sub-image, extracting contour information of the at least one sub-image; and if the size of the at least one sub-image is greater than or equal to the preset threshold, extracting the at least one sub-image Detailed information; wherein the image information includes the outline information and the detailed information.
  • a storage medium comprising a stored program, wherein the program executes an image information processing method having any of the above features.
  • a processor for executing a program, wherein the program executes an image information processing method having any of the above features.
  • the image to be processed is acquired; the image to be processed is converted into at least one sub image; and the image information of the at least one sub image is extracted according to the size of the at least one sub image, by converting the image to be processed into At least one sub-image of different sizes, and then extracting image information according to the size of the sub-image, achieves the purpose of different depth convolution of different size images, thereby achieving the technical effect of reducing the hardware consumption cost, thereby solving the problem of The technical problem of high hardware cost caused by convolution depth.
  • FIG. 1 is a schematic flowchart diagram of an optional image information processing method according to an embodiment of the present application
  • FIG. 2 is a schematic diagram of an optional image information processing method according to an embodiment of the present application.
  • FIG. 3 is a schematic structural diagram of an optional image information processing apparatus according to an embodiment of the present application.
  • FIG. 1 is an image information processing method according to an embodiment of the present application. As shown in FIG. 1 , the method includes the following steps:
  • Step S102 acquiring an image to be processed.
  • step S102 of the present application generally, for the neural network convolution, the input of the image is divided into different regions, and the image information is gradually extracted by convolving the different localities to input the maximum.
  • the bandwidth is 8 pixel data
  • 8 ⁇ 3 convolution units are used to perform 1 ⁇ 1 convolution operation on 8 pixel data
  • only 1 multiplier in each 3 ⁇ 3 convolution unit is substantially rolled.
  • the product operation so other multipliers are not used, which causes a waste of hardware resources.
  • the image information processing method of the present embodiment by extracting information of images of different resolution levels, a piece of original N-level pyramid can be gradually transformed into N small images of different sizes, and then different.
  • the small image performs sample extraction (image position, image information) of the two-dimensional information, and improves the usage rate of the convolution unit.
  • Step S104 converting the image to be processed into at least one sub image.
  • step S104 of the present application after the image to be processed is acquired, the image to be processed is converted into at least one sub-image.
  • the converting the image to be processed into the at least one sub-image comprises: reducing the image to be processed by N level, and generating N sub-images of different sizes.
  • the method for N-level reduction of the image to be processed may include reducing the processed image by N times year-on-year to obtain N different size sub-images.
  • Step S106 extracting image information of at least one sub-image according to the size of the at least one sub-image.
  • step S106 of the present application by extracting information of images of different resolution levels, by transforming a piece of original N-level pyramid, it is possible to gradually transform into N different-sized sub-images, and then to different sub-images.
  • sample extraction image position, image information
  • the contour information is extracted for the small-sized sub-image, the outline of the image can be eliminated for the large-sized sub-image, and the details are retained.
  • the same type of two-dimensional sampling refers to a method of recording and mapping the position and detail information of the current pixel point.
  • the same type of two-dimensional sampling generally different resolution images contain different details or contour information, for large-sized images (ie, large-resolution images) detailed information, and for small-sized images (ie, small-resolution images)
  • General outline information is more comprehensive, such as leaves
  • large-resolution images are generally clearer for the leaf details of the leaves
  • small-resolution images contain more information about the outline of the leaves.
  • Images for different resolutions can be stored by sampling the image details to generate a two-dimensional function f(x, y), where x, y represent the image position and f(x, y) represents the detail information.
  • extracting image information of the at least one sub-image according to the size of the at least one sub-image includes: determining whether the size of the at least one sub-image is less than a preset threshold; if the size of the at least one sub-image is smaller than the pre- And the threshold information is used to extract the contour information of the at least one sub-image; if the size of the at least one sub-image is greater than or equal to the preset threshold, the detailed information of the at least one sub-image is extracted; wherein the image information includes the contour information and the detail information.
  • extracting the detailed information of the at least one sub-image includes: extracting the image information of the at least one sub-image by using the same type of two-dimensional sampling; and extracting the contour information of the at least one sub-image from the image information to obtain the detailed information.
  • the culling the contour information of the at least one sub-image from the image information comprises: culling the contour information of the at least one sub-image from the image information by the filter, and locating the position and the detail of each pixel point of the at least one sub-image of the culling information Information is recorded and mapped.
  • the image information of the (N-1)th sub-image may be deleted to obtain the layer image information.
  • different sizes of sub-images carry different information.
  • small-sized sub-images will carry large outline information of images, while large-sized sub-images will be more More details of the image are carried, and different detail segments are obtained by extracting the sub-images of different sizes, and then neural network convolution is performed on different detail segments to realize image extraction.
  • the contour information includes at least one of the following: a shape feature and a spatial location feature; the detail information includes at least one of the following: a color feature, a texture feature.
  • the image to be processed is converted into at least one sub-image of different sizes, and then the image information is extracted according to the size of the sub-image, thereby achieving the purpose of different depth convolution for different size images, thereby achieving a reduction in hardware consumption cost.
  • the technical effect further solves the technical problem of high hardware cost due to increased convolution depth.
  • the method further includes: performing convolution operation on the image information of the at least one sub-image to output at least one convolution Result; the respective convolution results are summed to output the aggregated result.
  • the summation processing unit performs a step of summing the respective convolution results (for example, adding and summing) to output a total result.
  • performing convolution operation on the image information of the at least one sub-image to output the at least one convolution result comprises: performing convolution operation on the image information of the at least one sub-image by the plurality of convolution operation modules to output at least A convolution result.
  • the multiple convolution operation modules can process the image information of at least one sub-image in parallel, which has the purpose of improving processing efficiency and improving performance. Image information of different sub-images may be processed by different convolution operation modules.
  • An embodiment of the present application further provides an image information processing apparatus. It should be noted that the image information processing apparatus of this embodiment can be used to execute the image information processing method of the embodiment of the present application.
  • FIG. 3 is a schematic diagram of an image information processing apparatus according to an embodiment of the present application.
  • the upper computer includes an acquisition unit 20, a conversion unit 22, and an extraction unit 24.
  • An obtaining unit 20 configured to acquire an image to be processed
  • a converting unit 22 configured to convert the image to be processed into at least one sub-image
  • the extracting unit 24 is configured to extract image information of the at least one sub image according to the size of the at least one sub image.
  • the extracting unit 24 includes: a determining module, configured to determine whether a size of the at least one sub-image is smaller than a preset threshold; and an extracting module, configured to extract at least one sub-image if the size of the at least one sub-image is smaller than a preset threshold The contour information; if the size of the at least one sub-image is greater than or equal to a preset threshold, extracting the detail information of the at least one sub-image; wherein the image information includes the contour information and the detail information.
  • the extracting module is configured to: extract detailed information of the at least one sub-image by extracting image information of the at least one sub-image by using the same type of two-dimensional sampling; and extracting contour information of the at least one sub-image from the image information, to Get the details.
  • the image information processing apparatus further includes: a processing unit, configured to perform convolution operation on the image information of the at least one sub image to output at least one convolution result; and totalize the respective convolution results to output the total result.
  • a processing unit configured to perform convolution operation on the image information of the at least one sub image to output at least one convolution result; and totalize the respective convolution results to output the total result.
  • the processing unit includes: a plurality of convolution operation modules, configured to respectively perform convolution operations on the image information of the at least one sub-image to output at least one convolution result.
  • the image to be processed is acquired; the image to be processed is converted into at least one sub image; and the image information of the at least one sub image is extracted according to the size of the at least one sub image, by converting the image to be processed into At least one sub-image of different sizes, and then extracting image information according to the size of the sub-image, achieves the purpose of different depth convolution of different size images, thereby achieving the technical effect of reducing the hardware consumption cost, thereby solving the problem of The technical problem of high hardware cost caused by convolution depth.
  • An embodiment of the present application further provides an image information processing apparatus.
  • the image information processing apparatus includes: an image generator for acquiring an image to be processed; converting the image to be processed into at least one sub-image; and a processor executing a program, wherein the program is running for data output from the image generator The processing step of extracting image information of at least one sub-image according to the size of the at least one sub-image is performed.
  • the processor is configured to: according to the size of the at least one sub-image, extract image information of the at least one sub-image: determining whether a size of the at least one sub-image is smaller than a preset threshold; Extracting, according to the preset threshold, a size of the at least one sub-image, extracting contour information of the at least one sub-image; and if the size of the at least one sub-image is greater than or equal to the preset threshold, extracting the at least one sub-image Detailed information; wherein the image information includes the outline information and the detailed information.
  • the embodiment of the present application further provides a storage medium, where the storage medium includes a stored program, wherein the program executes an image information processing method having any of the above features.
  • the embodiment of the present application further provides a processor, where the processor is configured to run a program, where the program executes an image information processing method having any of the above features.
  • the disclosed technical contents may be implemented in other manners.
  • the device embodiments described above are only schematic.
  • the division of the unit may be a logical function division.
  • there may be another division manner for example, multiple units or components may be combined or may be Integrate into another system, or some features can be ignored or not executed.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, unit or module, and may be electrical or otherwise.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
  • the above integrated unit can be implemented in the form of hardware or in the form of a software functional unit.
  • the integrated unit if implemented in the form of a software functional unit and sold or used as a standalone product, may be stored in a computer readable storage medium.
  • a computer readable storage medium A number of instructions are included to cause a computer device (which may be a personal computer, server or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present application.
  • the foregoing storage medium includes: a U disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic disk, or an optical disk, and the like. .

Landscapes

  • Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biodiversity & Conservation Biology (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)

Abstract

本申请公开了一种图像信息处理方法及装置。其中，该方法包括：获取待处理图像；将所述待处理图像转换为至少一个子图像；根据所述至少一个子图像的尺寸，提取所述至少一个子图像的图像信息。本申请解决了由于加大卷积深度造成的消耗硬件成本较高的技术问题。

Description

图像信息处理方法及装置 技术领域
本申请涉及图像处理领域，具体而言，涉及一种图像信息处理方法及装置。
背景技术
目前随着人工智能的火热发展，对于图像的处理需求正在逐步的增大。如何快速的处理计算图像，更小误差的还原图像特征成为了目前的热点和难点。
目前大部分厂商通过逐步的加大卷积深度来获取图像更加准确的特征，然而，卷积深度越高，精确度越高，消耗的硬件成本就越高。
针对上述的问题，目前尚未提出有效的解决方案。
发明内容
本申请实施例提供了一种图像信息处理方法及装置，以至少解决由于加大卷积深度造成的消耗硬件成本较高的技术问题。
根据本申请实施例的一个方面，提供了一种图像信息处理方法，包括：获取待处理图像；将上述待处理图像转换为至少一个子图像；根据上述至少一个子图像的尺寸，提取上述至少一个子图像的图像信息。
可选地，上述根据上述至少一个子图像的尺寸，提取上述至少一个子图像的图像信息，包括：判断上述至少一个子图像的尺寸是否小于预设阈值；若上述至少一个子图像的尺寸小于上述预设阈值，提取上述至少一个子图像的轮廓信息；若上述至少一个子图像的尺寸大于等于上述预设阈值，提取上述至少一个子图像的细节信息；其中，上述图像信息包括上述轮廓信息和上述细节信息。
可选地，上述提取上述至少一个子图像的细节信息包括：采用同类二维采样提取上述至少一个子图像的图像信息；从上述图像信息中剔除上述至少一个子图像的轮廓信息，以得到细节信息。
可选地，根据上述至少一个子图像的尺寸，提取上述至少一个子图像的图像信息之后，上述方法还包括：对上述至少一个子图像的图像信息进行卷积运算，以输出至 少一个卷积结果；将各个卷积结果加总处理，以输出加总结果。
可选地，对上述至少一个子图像的图像信息进行卷积运算，以输出至少一个卷积结果包括：由多个卷积运算模块分别对上述至少一个子图像的图像信息进行卷积运算，以输出上述至少一个卷积结果。
根据本申请实施例的另一方面，还提供了一种图像信息处理装置，包括：获取单元，用于获取待处理图像；转换单元，用于将上述待处理图像转换为至少一个子图像；提取单元，用于根据上述至少一个子图像的尺寸，提取上述至少一个子图像的图像信息。
可选地，上述提取单元包括：判断模块，用于判断上述至少一个子图像的尺寸是否小于预设阈值；提取模块，用于若上述至少一个子图像的尺寸小于上述预设阈值，提取上述至少一个子图像的轮廓信息；若上述至少一个子图像的尺寸大于等于上述预设阈值，提取上述至少一个子图像的细节信息；其中，上述图像信息包括上述轮廓信息和上述细节信息。
根据本申请实施例的另一方面，还提供了一种图像信息处理装置，包括：图像生成器，用于获取待处理图像；将上述待处理图像转换为至少一个子图像；处理器，上述处理器运行程序，其中，上述程序运行时对于从上述图像生成器输出的数据执行如下处理步骤：根据上述至少一个子图像的尺寸，提取上述至少一个子图像的图像信息。
可选地，所述处理器用于执行以下步骤根据所述至少一个子图像的尺寸，提取所述至少一个子图像的图像信息：判断所述至少一个子图像的尺寸是否小于预设阈值；若所述至少一个子图像的尺寸小于所述预设阈值，提取所述至少一个子图像的轮廓信息；若所述至少一个子图像的尺寸大于等于所述预设阈值，提取所述至少一个子图像的细节信息；其中，所述图像信息包括所述轮廓信息和所述细节信息。
根据本申请实施例的另一方面，还提供了一种存储介质，上述存储介质包括存储的程序，其中，上述程序执行具有上述任意特征的图像信息处理方法。
根据本申请实施例的另一方面，还提供了一种处理器，上述处理器用于运行程序，其中，上述程序执行具有上述任意特征的图像信息处理方法。
在本申请实施例中，采用获取待处理图像；将待处理图像转换为至少一个子图像；根据至少一个子图像的尺寸，提取至少一个子图像的图像信息的方式，通过将待处理图像转换成尺寸不相同的至少一个子图像，进而根据子图像的尺寸提取图像信息，达到了对不同尺寸图像进行不同深度卷积的目的，从而实现了降低硬件消耗成本的技术效果，进而解决了由于加大卷积深度造成的消耗硬件成本较高的技术问题。
附图说明
此处所说明的附图用来提供对本申请的进一步理解，构成本申请的一部分，本申请的示意性实施例及其说明用于解释本申请，并不构成对本申请的不当限定。在附图中：
图1是根据本申请实施例的一种可选的图像信息处理方法的流程示意图；
图2是根据本申请实施例的一种可选的图像信息处理方法的示意图；
图3是根据本申请实施例的一种可选的图像信息处理装置的结构示意图。
具体实施方式
为了使本技术领域的人员更好地理解本申请方案，下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分的实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都应当属于本申请保护的范围。
需要说明的是，本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的本申请的实施例能够以除了在这里图示或描述的那些以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
实施例1
根据本申请实施例，提供了一种图像信息处理方法的方法实施例，需要说明的是，在附图的流程图示出的步骤可以在诸如一组计算机可执行指令的计算机系统中执行，并且，虽然在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤。
图1是根据本申请实施例的图像信息处理方法，如图1所示，该方法包括如下步骤：
步骤S102，获取待处理图像。
本申请上述步骤S102中，通常对于神经网络卷积来说，图像的输入会被分割划分为不同的区域，通过对不同的局域进行卷积来逐渐的实现对图像信息的提取，以输入最大带宽为8个画素数据为例，当使用8个3×3卷积单元对8个画素数据进行1×1卷积运算时，各3×3卷积单元中仅1个乘法器实质上进行卷积运算，因此其他的乘法器没有运用到，这造成硬件资源浪费。而本实施例的图像信息处理方法，通过将不同分辨率层次的图像进行信息提取，通过将一张原始经过N等级的金字塔变化，可以逐渐的变换成N个不同尺寸的小图像，然后对不同的小图像进行二维信息的采样提取(图像位置，图像信息)，提升卷积单元的使用率。
步骤S104，将待处理图像转换为至少一个子图像。
本申请上述步骤S104中，在获取到待处理图像之后，将待处理图像转换为至少一个子图像。其中，将待处理图像转换为至少一个子图像，包括：将待处理图像进行N等级缩小，生成N个不同尺寸的子图像。
其中，将待处理图像进行N等级缩小的方法可以包括将处理图像进行同比缩小N次，得到N个不同尺寸的子图像。
步骤S106，根据至少一个子图像的尺寸，提取至少一个子图像的图像信息。
本申请上述步骤S106中，通过将不同分辨率层次的图像进行信息提取，通过将一张原始经过N等级的金字塔变化，可以逐渐的变换成N个不同尺寸的子图像，然后对不同的子图像进行二维信息的采样提取(图像位置，图像信息)，由于对小尺寸的子图像进行了轮廓信息提取，在对于大尺寸的子图像中可以将图像的轮廓进行剔除，而对细节进行保留，对大尺寸的子图像进行同类二维采样提取。本实施例中，同类二维采样是指对当前像素点的位置与细节信息进行记录、映射的方法。
具体地，同类二维采样：一般不同分辨率的图像所包含的细节信息或轮廓信息是不同的，对于大尺寸的图像(即大分辨率的图像)细节信息比较多，而对于小尺寸的图像(即小分辨率的图像)一般轮廓信息比较全面，比如叶子，大分辨率的图像一般对于叶子的脉络细节比较清楚，而小分辨率的图像对叶子的轮廓含有的信息比较多。对于不同的分辨率的图像可以通过对图像细节采样来生成一个二维的函数f(x,y)进行存储，其中x,y代表图像位置，f(x,y)代表细节信息。
作为一种可选的实现方式，根据至少一个子图像的尺寸，提取至少一个子图像的图像信息，包括：判断至少一个子图像的尺寸是否小于预设阈值；若至少一个子图像的尺寸小于预设阈值，提取至少一个子图像的轮廓信息；若至少一个子图像的尺寸大于等于预设阈值，提取至少一个子图像的细节信息；其中，图像信息包括轮廓信息和 细节信息。
可选地，提取至少一个子图像的细节信息包括：采用同类二维采样提取所述至少一个子图像的图像信息；从图像信息中剔除至少一个子图像的轮廓信息，以得到细节信息。
其中，从图像信息中剔除至少一个子图像的轮廓信息包括：通过滤波器从图像信息中剔除至少一个子图像的轮廓信息，并且对剔除轮廓信息的至少一个子图像的各个像素点的位置与细节信息进行记录、映射。
如图2所示，在图像信息提取时，对于第N层子图像，可以删除第N-1层子图像的图像信息，得到该层图像信息。对于不同尺寸的子图像上而言，不同的尺寸大小的子图像所携带的信息不同，一般而言，小尺寸的子图像会携带图像的大的轮廓信息，而大尺寸的子图像则会更多的携带图像的细节信息，通过对不同尺寸的子图像进行细节提取会初步得到不同的细节片段，然后对不同的细节片段进行神经网络卷积实现图像提取。
可选地，轮廓信息包括以下至少之一：形状特征和空间位置特征；细节信息包括以下至少之一：颜色特征、纹理特征。
通过上述步骤，将待处理图像转换成尺寸不相同的至少一个子图像，进而根据子图像的尺寸提取图像信息，达到了对不同尺寸图像进行不同深度卷积的目的，从而实现了降低硬件消耗成本的技术效果，进而解决了由于加大卷积深度造成的消耗硬件成本较高的技术问题。
作为一种可选的实现方式，根据至少一个子图像的尺寸，提取至少一个子图像的图像信息之后，方法还包括：对至少一个子图像的图像信息进行卷积运算，以输出至少一个卷积结果；将各个卷积结果加总处理，以输出加总结果。
其中，由加总处理单元执行将各个卷积结果加总处理(例如是相加加总)，以输出加总结果的步骤。
可选地，对至少一个子图像的图像信息进行卷积运算，以输出至少一个卷积结果包括：由多个卷积运算模块分别对至少一个子图像的图像信息进行卷积运算，以输出至少一个卷积结果。多个卷积运算模块可以平行处理至少一个子图像的图像信息，已达到提高处理效率，增进效能的目的。不同的子图像的图像信息可以由不同的卷积运算模块处理。
需要说明的是，在附图的流程图示出的步骤可以在诸如一组计算机可执行指令的 计算机系统中执行，并且，虽然在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤。
实施例2
本申请实施例还提供了一种图像信息处理装置。需要说明的是，该实施例的图像信息处理装置可以用于执行本申请实施例的图像信息处理方法。
图3是根据本申请实施例的一种图像信息处理装置的示意图。如图3所示，该上位机包括：获取单元20、转化单元22和提取单元24。
获取单元20，用于获取待处理图像；
转换单元22，用于将待处理图像转换为至少一个子图像；
提取单元24，用于根据至少一个子图像的尺寸，提取至少一个子图像的图像信息。
可选地，提取单元24包括：判断模块，用于判断至少一个子图像的尺寸是否小于预设阈值；提取模块，用于若至少一个子图像的尺寸小于预设阈值，提取至少一个子图像的轮廓信息；若至少一个子图像的尺寸大于等于预设阈值，提取至少一个子图像的细节信息；其中，图像信息包括轮廓信息和细节信息。
可选地，提取模块用于执行以下步骤提取至少一个子图像的细节信息：采用同类二维采样提取所述至少一个子图像的图像信息；从图像信息中剔除至少一个子图像的轮廓信息，以得到细节信息。
可选地，图像信息处理装置还包括：处理单元，用于对至少一个子图像的图像信息进行卷积运算，以输出至少一个卷积结果；将各个卷积结果加总处理，以输出加总结果。
可选地，处理单元包括：多个卷积运算模块，用于分别对至少一个子图像的图像信息进行卷积运算，以输出至少一个卷积结果。
在本申请实施例中，采用获取待处理图像；将待处理图像转换为至少一个子图像；根据至少一个子图像的尺寸，提取至少一个子图像的图像信息的方式，通过将待处理图像转换成尺寸不相同的至少一个子图像，进而根据子图像的尺寸提取图像信息，达到了对不同尺寸图像进行不同深度卷积的目的，从而实现了降低硬件消耗成本的技术效果，进而解决了由于加大卷积深度造成的消耗硬件成本较高的技术问题。
实施例3
本申请实施例还提供了一种图像信息处理装置。该图像信息处理装置包括：图像生成器，用于获取待处理图像；将待处理图像转换为至少一个子图像；处理器，处理器运行程序，其中，程序运行时对于从图像生成器输出的数据执行如下处理步骤：根据至少一个子图像的尺寸，提取至少一个子图像的图像信息。
可选地，所述处理器用于执行以下步骤根据所述至少一个子图像的尺寸，提取所述至少一个子图像的图像信息：判断所述至少一个子图像的尺寸是否小于预设阈值；若所述至少一个子图像的尺寸小于所述预设阈值，提取所述至少一个子图像的轮廓信息；若所述至少一个子图像的尺寸大于等于所述预设阈值，提取所述至少一个子图像的细节信息；其中，所述图像信息包括所述轮廓信息和所述细节信息。
本申请实施例还提供了一种存储介质，存储介质包括存储的程序，其中，程序执行具有上述任意特征的图像信息处理方法。
本申请实施例还提供了一种处理器，处理器用于运行程序，其中，程序执行具有上述任意特征的图像信息处理方法。
上述本申请实施例序号仅仅为了描述，不代表实施例的优劣。
在本申请的上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述的部分，可以参见其他实施例的相关描述。
在本申请所提供的几个实施例中，应该理解到，所揭露的技术内容，可通过其它的方式实现。其中，以上所描述的装置实施例仅仅是示意性的，例如所述单元的划分，可以为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，单元或模块的间接耦合或通信连接，可以是电性或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以 是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
所述集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可为个人计算机、服务器或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、移动硬盘、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述仅是本申请的优选实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本申请原理的前提下，还可以做出若干改进和润饰，这些改进和润饰也应视为本申请的保护范围。

Claims (11)

  1. 一种图像信息处理方法，包括：
    获取待处理图像；
    将所述待处理图像转换为至少一个子图像；
    根据所述至少一个子图像的尺寸，提取所述至少一个子图像的图像信息。
  2. 根据权利要求1所述的方法，其中，所述根据所述至少一个子图像的尺寸，提取所述至少一个子图像的图像信息，包括：
    判断所述至少一个子图像的尺寸是否小于预设阈值；
    若所述至少一个子图像的尺寸小于所述预设阈值，提取所述至少一个子图像的轮廓信息；
    若所述至少一个子图像的尺寸大于等于所述预设阈值，提取所述至少一个子图像的细节信息；
    其中，所述图像信息包括所述轮廓信息和所述细节信息。
  3. 根据权利要求2所述的方法，其中，所述提取所述至少一个子图像的细节信息包括：
    采用同类二维采样提取所述至少一个子图像的图像信息；
    从所述图像信息中剔除所述至少一个子图像的轮廓信息，以得到所述细节信息。
  4. 根据权利要求1所述的方法，其中，根据所述至少一个子图像的尺寸，提取所述至少一个子图像的图像信息之后，所述方法还包括：
    对所述至少一个子图像的图像信息进行卷积运算，以输出至少一个卷积结果；
    将各个卷积结果加总处理，以输出加总结果。
  5. 根据权利要求4所述的方法，其中，对所述至少一个子图像的图像信息进行卷积运算，以输出至少一个卷积结果包括：
    由多个卷积运算模块分别对所述至少一个子图像的图像信息进行卷积运算，以输出所述至少一个卷积结果。
  6. 一种图像信息处理装置，包括：
    获取单元，用于获取待处理图像；
    转换单元，用于将所述待处理图像转换为至少一个子图像；
    提取单元，用于根据所述至少一个子图像的尺寸，提取所述至少一个子图像的图像信息。
  7. 根据权利要求6所述的装置，其中，所述提取单元包括：
    判断模块，用于判断所述至少一个子图像的尺寸是否小于预设阈值；
    提取模块，用于若所述至少一个子图像的尺寸小于所述预设阈值，提取所述至少一个子图像的轮廓信息；若所述至少一个子图像的尺寸大于等于所述预设阈值，提取所述至少一个子图像的细节信息；
    其中，所述图像信息包括所述轮廓信息和所述细节信息。
  8. 一种图像信息处理装置，包括：
    图像生成器，用于获取待处理图像；将所述待处理图像转换为至少一个子图像；
    处理器，所述处理器运行程序，其中，所述程序运行时对于从所述图像生成器输出的数据执行如下处理步骤：根据所述至少一个子图像的尺寸，提取所述至少一个子图像的图像信息。
  9. 根据权利要求8所述的装置，其中，所述处理器用于执行以下步骤根据所述至少一个子图像的尺寸，提取所述至少一个子图像的图像信息：
    判断所述至少一个子图像的尺寸是否小于预设阈值；
    若所述至少一个子图像的尺寸小于所述预设阈值，提取所述至少一个子图像的轮廓信息；
    若所述至少一个子图像的尺寸大于等于所述预设阈值，提取所述至少一个子图像的细节信息；
    其中，所述图像信息包括所述轮廓信息和所述细节信息。
  10. 一种存储介质，所述存储介质包括存储的程序，其中，所述程序执行权利要求1至5中任意一项所述的图像信息处理方法。
  11. 一种处理器，所述处理器用于运行程序，其中，所述程序运行时执行权利要求1至5中任意一项所述的图像信息处理方法。
PCT/CN2018/099598 2017-10-19 2018-08-09 图像信息处理方法及装置 Ceased WO2019076110A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710989393.7 2017-10-19
CN201710989393.7A CN107729905A (zh) 2017-10-19 2017-10-19 图像信息处理方法及装置

Publications (1)

Publication Number Publication Date
WO2019076110A1 true WO2019076110A1 (zh) 2019-04-25

Family

ID=61213267

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/099598 Ceased WO2019076110A1 (zh) 2017-10-19 2018-08-09 图像信息处理方法及装置

Country Status (2)

Country Link
CN (1) CN107729905A (zh)
WO (1) WO2019076110A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107729905A (zh) * 2017-10-19 2018-02-23 珠海格力电器股份有限公司 图像信息处理方法及装置
CN110555847B (zh) * 2019-07-31 2021-04-02 瀚博半导体(上海)有限公司 一种基于卷积神经网络的图像处理方法及装置

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040046768A1 (en) * 2000-05-10 2004-03-11 Nec Corporation Method, system and record medium for generating wide-area high-resolution image
CN101882305A (zh) * 2010-06-30 2010-11-10 中山大学 一种图像增强处理的方法
CN106296600A (zh) * 2016-08-01 2017-01-04 大连理工大学 一种基于小波变换图像分解的对比度增强方法
CN107729905A (zh) * 2017-10-19 2018-02-23 珠海格力电器股份有限公司 图像信息处理方法及装置

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103218831B (zh) * 2013-04-21 2015-11-18 北京航空航天大学 一种基于轮廓约束的视频运动目标分类识别方法
CN104835130A (zh) * 2015-04-17 2015-08-12 北京联合大学 一种多曝光图像融合方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040046768A1 (en) * 2000-05-10 2004-03-11 Nec Corporation Method, system and record medium for generating wide-area high-resolution image
CN101882305A (zh) * 2010-06-30 2010-11-10 中山大学 一种图像增强处理的方法
CN106296600A (zh) * 2016-08-01 2017-01-04 大连理工大学 一种基于小波变换图像分解的对比度增强方法
CN107729905A (zh) * 2017-10-19 2018-02-23 珠海格力电器股份有限公司 图像信息处理方法及装置

Also Published As

Publication number Publication date
CN107729905A (zh) 2018-02-23

Similar Documents

Publication Publication Date Title
Cai et al. Retinexformer: One-stage retinex-based transformer for low-light image enhancement
Wu et al. Fast end-to-end trainable guided filter
Long et al. Bishift networks for thick cloud removal with multitemporal remote sensing images
Afifi et al. Cie xyz net: Unprocessing images for low-level computer vision tasks
US10853916B2 (en) Convolution deconvolution neural network method and system
CN109146788B (zh) 基于深度学习的超分辨率图像重建方法和装置
CN111709883A (zh) 一种图像检测方法、装置及设备
CN106127688B (zh) 一种超分辨率图像重建方法及其系统
CN111127374A (zh) 一种基于多尺度密集网络的Pan-sharpening方法
CN105046651B (zh) 一种图像的超分辨率重构方法和装置
CN115439325B (zh) 低分辨率高光谱图像处理方法、装置、计算机程序产品
CN110418139B (zh) 一种视频超分辨修复方法、装置、设备及存储介质
CN114742733A (zh) 云去除方法、装置、计算机设备和存储介质
WO2019076109A1 (zh) 图像信息的池化方法及装置、存储介质、处理器
CN104346630B (zh) 一种异构特征融合的云花卉识别方法
CN113869330A (zh) 水下鱼类目标检测方法、装置及存储介质
CN105741243A (zh) 一种模糊图像复原方法
CN105550989A (zh) 基于非局部高斯过程回归的图像超分辨方法
CN107103585A (zh) 一种图像超分辨率系统
CN101739670A (zh) 非局部均值空域时变图像滤波方法
WO2019076110A1 (zh) 图像信息处理方法及装置
CN114049491A (zh) 指纹分割模型训练、指纹分割方法、装置、设备及介质
Fang et al. Learning explicit smoothing kernels for joint image filtering
Zhang et al. Clouds and haze co-removal based on weight-tuned overlap refinement diffusion model for remote sensing images
CN111931793B (zh) 一种显著性目标提取方法及系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18869082

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 09/10/2020)

122 Ep: pct application non-entry in european phase

Ref document number: 18869082

Country of ref document: EP

Kind code of ref document: A1